Norbu Project Elizaveta Gorbunova

Table of contents

Project's goal

Investigate app usage and provide recommendations which will allow the Norbu app increase its Retention rate on 28 day from 4% to 20%.

Project's summary

What has been done:

Client's questions:

  1. Sleep problems

What proportion of those participating in the test survey have sleep problems, in which countries, cities? What does the Device model affect

  1. LTV

Calculate user LTV all Calculate user LTV / GEO Calculate user LTV / Traffic source (medium) Calculate user LTV / Platform Calculate user LTV revenue What affects LTV?

  1. Retention

Retention 1,3,7,30 day / Platform Retention 1,3,7,30 day / GEO What affects Retention?

  1. Churn

What affects the removal (app remove)?

  1. Purchases

What affects the in-app purchase / product ID What affects the renewal of the subscriptions? What affects user engagment?

  1. User journey

Derive and analyze patterns of User journey behavior Correlations Stress evaluation / Pulse-stress-test

How clear and easy to learn is the MBSC mindfulness meditation level 1

% started 5-day-unlock a traning % completed 5-day-unlock a traning % user started 5-day-unlock a traning -> not completed this unlock -> made purchase of this training

Resources

  1. https://www.jmir.org/2019/9/e14567/PDF
  1. https://www.theneura.com/headspace-mobile-engagement-strategy/ (more of an ad)
  1. https://www.businessofapps.com/data/calm-statistics/

Data preprocessing

  1. Connect to BigQuery to collect Norbu app data
  2. Load data from Google cloud for a faster usage
  3. Preprocess data:
    • leave only relevant data
    • remove duplicates
    • rename columns
    • handle Null values
    • change data types
    • expand data saved in a dictionary

Connect to BigQuery

Query data

Load data and save locally for faster use

Events

Users

Devices

Geography

Traffic

Preprocess data

Events

Users

Devices

Let's extract relevant data from 'device' column

Geo

Traffic

Events

User scenarios/Event funnels

Let's define several most common user scenarios and investigate those.

  1. scr_home - user is on a page with workouts
  2. scr_training_details - user views training details

Then different trainings:

scr_ballshome scr_ballsgame result_game

select_training_qh0 - Manage your stress start_training_qh0

? scr_intro2 User was on screen intro 2 intro_next_btn_tapped on intro2 the user clicked on the 'continue' button

After trainings user should get to the page with a survey "What you feel?": scr_whatyoufeel

  1. session_start
  2. scr_survey survey_start survey_end
    scr_survey_result

Types of events

Events share

Calculate the proportion of users who performed the action at least once

Breathing exercise

Balls game

Manage stress training (MBSC)

Events funnel

Events per week

Events per day

Users per day

Events per user

Number of users per day/week

Sticky factor

Cohort Analysis

Retention rate

LTV - TBD

Amount of revenue and in-app purchases differs, may be not all purchases are recorded.

Users churn

Let's investigate how many users stopped using the app. This will include users:

As we have just 2 months of data on users' events, let's try to define the period of inactivity after which we can consider the user to be lost. Let's check how much time passed from last event of each user.

Period of inactivity without returning to the app

From the graph we can notice that staring from 10 days of inactivity there is the same amount of inactive users, let's consider users, who are inactvie more than 10 days to be lost. Let's check how many inactive users we have right now:

Period of inactivity with returning back to the app

between one event and previous event - for how long people stop using the app and then return

We calculate for each user maximum period of inactivity: create a column 'time_between_events' and find maximum number:

We can notice that most users do not stop using the app for more than 10 days. There is a small amount of users, who return to the app after a long period of inactivity.

App removal

Share of users who remove the app is around 50% which corresponds to average uninstall rate in mobile app industry.

Let's combine users who were inactive for more than 10 days and those who removed an app to calculate the total number of churn users

Number of app removals over time

To understand why customers leave, first understand why customers stay

Most active users

Devices

Most of users use mobile phones

Geo

Treemap

Treemap regions

Traffic

There are just 2 significant traffic sources: direct and google play

Users actvity by country

Almost 50% of unique users come from Iran and Germany. Why?

There are some spikes and declines, but overall number of users is ditributed quite evenly.

Users actvity by source

There are just two traffic sources for now.

Conclusion - TBD